This topic focuses on EEG-based seizure prediction using an online system with group learning, adaptive postprocessing, and feature encoding, achieving high sensitivity and low falsepositive rates in long-term data (>6 months). It addresses seizure clustering in real-world scenarios and demonstrates the effectiveness of this novel approach. The second topic about “unsupervised EEG feature Learning for Seizure Prediction based on convolutional Autoencoder” explores unsupervised feature learning using convolutional autoencoders (CAEs) applied to the CHB-MIT scalp EEG database for better prediction.
Research should explore advanced deep learning models, expand datasets for diverse patient data, and integrate real-time adaptive strategies with additional physiological signals to improve prediction accuracy.
Author: PhD student: Vidhatri Gujela